Assessing the clinical importance of symptomatic improvements. An illustration in rheumatology

Abstract

Objective: To estimate when a difference in disability symptoms is sufficiently large to be important to individual patients.

Design: Cross-sectional analysis of two groups: derivation set (n = 46) and validation set (n = 57).

Setting: The Arthritis Foundation, Northern California Chapters.

Participants: Volunteer sample of patients with arthritis who live in the community.

Main outcome measures: We applied the Stanford Health Assessment Questionnaire to assess the functional status of individuals. Participants then conducted one-on-one conversations with each other and rated whether their disability was "much better" "somewhat better," "about the same," "somewhat worse," or "much worse" relative to each person they met. For every conversation we calculated the difference between the two participants' health assessment questionnaire scores and linked the difference to the subjective comparison ratings of each individual in the pair.

Results: Health assessment questionnaire score differences were significantly correlated with subjective comparison ratings (correlation coefficient, .41; 95% confidence interval, 0.31 to 0.50). We estimated that health assessment questionnaire scores needed to differ by about 0.19 units for average respondents to stop rating themselves as "about the same" and start rating themselves as "somewhat better" (95% confidence interval, 0.10 to 0.28). Analysis of a second group of patients revealed a similar threshold (mean, 0.23 units; 95% confidence interval, 0.13 to 0.23). In both groups, health assessment questionnaire score differences were imperfect predictors of individual ratings and the threshold for less disabled participants tended to be lower than the threshold for more disabled participants.

Conclusions: Some statistically significant differences in functional status scores may be so small that they represent trivial degrees of symptom relief. An awareness of the smallest difference in symptom scores that is important to patients can provide a rough guide to help clinicians interpret the medical literature.